IT
Agents use your app in a real browser, on a laptop and a phone, and report the problems a customer would run into.
See how a first run worksWe test software the way people use it.
Our agents use your app like a stranger would and report what breaks, with evidence for each finding.
nth Labs is the testing team for people who build with AI.
Write to us
Software is now written faster than people can check it. nth Labs does the checking.
On the largest code host, output rose by a fifth to a quarter in a year, while every sign that someone went back to read the code fell. Both figures come from five published studies, summarized in our post.
Software production is booming. Who is checking it?What we build
nth Labs is for people who build with AI tools and ship faster than they can test. There is nothing to install and no test to write.
Marketing
Flags the claims your pages cannot back up, and asks simulated buyers from your market whether they would try it.
How the panel picks its testers
Finance
Estimates what each fix is worth and what your first year could earn, based on where people give up.
Test your pricing page
Each finding comes with its evidence and a fix prompt for your coding tool. Once you fix it, the next run checks that the fix held.
Fix prompts and verifyingWe also publish comparisons with Octomind, Aaru and Synthetic Users, including when each one is the better choice.
How we work
01
We confirm findings before we report them
When a model says something is wrong, a real browser replays it before it can block a release. If the browser cannot confirm it, the finding is marked unverified.
How findings are confirmed
02
We only publish numbers we measured
Accuracy figures come from your verdicts on our findings. Where there is not enough data yet, we say so instead of estimating.
How accuracy is measured
03
Every regression check is tested first
A confirmed finding becomes a check that has to fail on the broken page before it counts. A check that has never failed is marked unproven.
How regression checks work
04
We publish our misses
We plant bugs in test apps and publish what nth Labs caught and what it missed, with the commands to reproduce each run.
Published resultsOur commitments
- No tracking or training
- This website sets no cookies, and the console sets only the first-party cookies it needs. We do not sell your data or use it to train models.
- No API keys to manage
- Checks that need a model run on our AI service as part of your plan. Checks that only measure call no model at all.
- Scores only for what we reached
- Every run lists the pages it could not reach and scores only the ones it did.
- Everything exports
- Settings exports every run and finding as one JSON file, and each report downloads as a single HTML file.
Where we are
In beta
The hosted console at console.nthlabs.dev is in beta, alongside a command line, a CI step and tools for coding agents.
A tour of the consoleFree to start
The free checks are unlimited on every plan, and the Free plan needs no card. Paid plans add validations, projects and people.
Changing quickly
The product changes week to week. A figure we have not measured yet is marked that way.
How accurate the panel isContact
hello@nthlabs.devBased in Toronto.